Address Matching Accident Records Using a Map Reference Coordinate System
Janet H. Gritzner
Address Matching Accident Records Using a Map Reference Coordinate System
State-recorded accident record data are often logged by mileage reference markers (MRMs) or by map reference coordinates. Using accident data with map reference coordinates in addressmatching applications require conversion of XY map locations to a route measure system.
The street center line coverage for Sioux Falls, South Dakota is geocoded with XY map references. Accident record XY map locations are converted to street names with address ranges. Normal address matching routines are used to locate along-the-street and intersection accidents. These data will be used in ArcView and ArcInfo for accident analysis, including the development of collision diagrams.
XY map locations are assigned to node intersections of the Sioux Falls street center line dataset through a series of coordinate transformations. Arc from-X and to-X and from-Y and to-Y map references are derived from the intersection coordinate values. Street-arcs are assigned X or Y constants and from-address and to-address ranges dependent on street orientation. North-south oriented streets have X constants with a from-Y and to-Y address range and east-west oriented streets have Y constants with a from-X to-X address range. Route systems are constructed based on street names with measures taken from the from-address and to-address attributes. Lookup tables are developed from the route section INFO table. These tables are used to give street names to accident records with no street information. Once accidents are located by street name, the accidents are matched to the street center line dataset through normal address matching procedures.
Reference Systems for Accident Data
A common method of logging accident data on a state wide level is by referencing accident locations by mileage reference markers (MRMs) or by map reference coordinates. The preferred method in rural sections of South Dakota is to use the MRMs that are placed along all state- administered highways to reference an accident location. A distance from an established MRM is normally part of official reporting of the accident. In South Dakota cities, the largest of which is Sioux Falls, accidents are normally located by map reference coordinates. The MRM system is not useful for locating accidents in urban settings in that MRMs are only established for the state-administered highways. Even when state highways bisect urban places, MRMs are rarely positioned so that they may be used by law-enforcement officers in reporting accident location. Attending officers normally use the names of intersecting streets as a locational reference. Most accidents happen at or near intersections. For non-intersection accidents, accident reports reference distances from intersecting streets.
In South Dakota, accidents with injury or monetary value over $500 are reported to the state. Accidents are detailed on standard accident reporting forms. The Accident Records office in Pierre receive accident reports from a variety of law-enforcement agencies including the South Dakota highway patrol, county sheriff offices and local police departments. Accident report review at the state-level involves assigning map reference coordinates to accidents and entering selected data elements into a database.
The map reference system used in South Dakota is county-based. The reference origin is in the northwest corner of the county. Public land survey section lines serve as primary references. XY coordinate pair 1.00 2.00 in Minnehaha County refers to the intersection of the first north-south section line and second east-west section line or roughly one mile by two miles from the corner of the county. Coordinate pair 22.75 21.50 in the county is a location three-fourths of the way between north-south section lines22 and 23 and half of the distance between east-west section lines 21 and 22. It equates to the intersection of Lowell Ave and 5th St. in Sioux Falls. The Accident Records office has manually assigned some 2600 coordinate pairs to intersecting streets in Sioux Falls. These coordinates are used to assign location to intersection accidents. The so-called book coordinates exist for about 40% of Sioux Falls intersections.
Accident record data delivered by the state office use county reference numbers and a map reference coordinate as the sole locational element. Street names are not included in the records. Using accident data with map reference coordinates in addressmatching applications require conversion of XY map locations to a route measure system including reassignment of street names.
Task Objectives
The project objectives are three-fold: The first objective or task is to attribute intersections of the Sioux Falls center line coverage with equivalent map reference coordinates, second objective is to geocode the street coverage with map reference coordinate ranges, and third to develop a address file from reported accidents in preparation for address matching.
Task 1 - Transfer map reference coordinates to geo-referencedstreet dataset
The first requirement is to attribute the intersections of the geo-referenced Sioux Falls street center line coverage. A route-measure system is developed for the section line streets crisscrossing the city. County-based map reference coordinates are used as the measure units. North-south streets have x-constants and east-west streets have y-constants.
Values from the f-meas and t-meas items on the section tables are used to create tics at node positions in an empty coverage. NODEPOINT is used to produce a point coverage from the node intersections of the street center line coverage. New tics are added to the point coverage at intersection positions of the section line streets. The point coverage is transformed to map reference coordinates with the transform function. ADDXY added true X coordinate and Y coordinate positions, which are use to verify the transformation.
The next step involved producing a point coverage of book coordinates. The NEAR function is used to assign the nearest book coordinate to the point coverage of node intersections now in map reference coordinates. A search distance of .026 is used to identify the nearest book coordinate. The book coordinates are then moved to the PAT of the point coverage of nodes. Book coordinates are visually verified and corrected as needed. A new tic coverage is generated using the assigned book coordinates. The point coverage of nodes is transformed into the tic coverage to create a closer approximation of map reference coordinates as recorded in Accident Record office. The process is repeated until all locatable book coordinates are identified. Some 2400 of 2600 book coordinates are assigned. For unassigned coordinates, there are five categories of error. They include: 1) no intersection, 2) missing node, 3) non-existent street, 4) duplicate entry, and 5) reversed entry. The software interpolates map coordinates for intersections with no book XY code or for some 2453 intersections.
ADDXY provides coordinate values for each point or intersection. The new coordinate information is added to the geo-reference point coverage of node intersections which is then joined to the Sioux Falls street center line NAT.
Task 2 - Geocode the street coverage with map reference coordinate ranges
An Esri AML st_name2nodes.aml moves street name information from the street arcs to nodes (i.e., from the street center line AAT to the NAT). Using a series of relates, from node and to node map reference coordinates are transferred from the NAT to items designated as from-X-address to to-X-address and from-Y-address toto-Y-address in the geo-referenced street center line AAT. Once transferred, the from and to X and Y ranges are sorted for correct direction. X coordinate ranges are sorted in ascending order from west to east and Y coordinate ranges are ordered north to south.
Street arcs are then assigned a X or Y constant on the basis of a statistical constant value. Generally avenues run north-south in Sioux Falls and streets east-west with some exceptions. Avenues have X constants and street Y constants. Other street types (drives, roads, circles, trails) including diagonal and circular streets are given constants based on statistical variation.
Ultimately all street arcs are given a value of X or Y. A from-address item, to-address item, from-constant item and to-constant item are added to the AAT. For X constant streets the Y map reference coordinate ranges are encoded as from- and to-addresses and X ranges as from- and to-constants. For Y-constant streets X map reference coordinate ranges are transferred to the from- and to addresses and Y ranges to the from- and to-constants. An ADD file is created and built using the from- and to-address and street name as the geo-coded address.
Task 3 - Develop an address file of reported accidents
Crucial to developing an address file of accidents is identifying the constant coordinate of the X coordinate and Y coordinate accident identifier in accident records. Once identified, a lookup table is used to assign a street name to the coordinate with the other X or Y coordinate being the address number element. An X coordinate of 22.00 and Y coordinate of 22.68 translates to 22.68 CLIFF AVE.
The look-up table is constructed by building route and section tables for the street center line AAT. The street names become routes and measures are taken from the from-constant and to-constant items with an added attribute labeling the street as to its X or Y orientation. A second set of routes and sections are built from the from-address and to address element. STATISTICS command produces tables of minimum and maximum measures. These are used in determining which coordinate of the X and Y coordinate items of the accident record is the constant map coordinate.
For constants with similar coordinates, a second lookup table with the minimum from address range and maximum to address range distinguishes one route or street from another. Based on four look-up tables (two for X constant streets and two for Y constant streets), street names and a map reference positions are assigned to the accidents. These become the components of a parsed address field in the accident records. Accident records are now in a form for normal addressmatching in ArcView, ArcInfo and PC ARC/INFO.
Concluding Remarks
The procedure of converting XY map locations to a route measure system while possible is not direct or obvious. It is a process that will be used in South Dakota until the state office adopts new formats for logging accident data. It is strongly recommended that the street name field be added to the data form to ease the building address files needed for address matching.
Janet H. Gritzner
_